recognize
recognize copied to clipboard
nextcloud
Include a Process to Review or Approve Newly Categorized Faces to an Existing Named Person/Cluster
Describe the feature you'd like to request
Due to the complexity of AI Facial Recognition, and understanding that it will never be perfect, it would be beneficial to have a Review or Approval process when newly categorized faces are added to an existing Named Person/Cluster.
For example. I have a Person that I have edited and renamed as "Brian" instead of the default Cluster ID # 123. This means I know and care about the Person. I have 3,000 images in the Person Cluster named Brian. When faces are detected, categorized as Brian, and added to the Brian cluster (ID # 123), it's very cumbersome and nearly impossible to find this image in the Brian cluster to be sure it was properly categorized. This is especially true if adding older images to Nextcloud, as verifying newer images would be a bit easier but still quite cumbersome.
This type of feature could also be beneficial to https://github.com/nextcloud/recognize/issues/606
Describe the solution you'd like
I want the option to enable a review and/or approval process when new faces are categorized into an already-established Person/Cluster.
Even something as simple as some type of report showing which images were added to which clusters would be helpful so they can be reviewed for accuracy.
Describe alternatives you've considered
Click on the Person I expect an image to be categorized as and carefully review the images to be sure they were categorized properly. If they're not found in the expected cluster, perform the manual review of other existing clusters. Honestly, there is no good alternative for this that I can think of.
I'll just leave this comment here as a reminder based on the work that's being done currently on the clustering algorithm:
If/when soft clustering is implemented it would make sense to automatically assign a new detection into an existing clusters if the probability of said detection being in the cluster is high (e.g. >90%). The detections with probability in the grey area would be shown to the user with a representative photo from the prospective cluster: "Are these the same person?" This way it would be easy to nudge the clustering algorithm to the right direction when it's unsure.
What happens to those photos at the moment? Are they recycled to be tagged later or is the tag completely removed?
We are currently maintaining this app on a limited effort basis. This means Nextcloud GmbH will not invest further development resources ourselves in advancing this app with new features. That doesn't mean there will be no new features, however: We do review and enthusiastically welcome community Pull Requests. We would be more than excited if you would like to collaborate with us on this issue. Feel free to reach out here in the comments if you would like to work on this and have questions about how to go about it, how to proceed or would like a short introduction call into the code base. I'm here to help with your questions :v: (See https://github.com/nextcloud/recognize/discussions/779 for more information on this)
This is sad, because it's one of the biggest selling points for me at least... I have over 43k photos to dive through and I'd love to have this as my go-to solution to keep track :)
